Bitaxe – 12V PSU Connected to 5V Board (Permanent Damage)

Unplug the wrong PSU at the wall, not the board. Some bench PSUs and cheap adapters keep output live for 2-5 seconds after AC is cut. Pull the plug from the outlet, then disconnect the barrel jack from the Bitaxe. Every second of 12V exposure extends the damage zone on the board, so this is a time-sensitive first move.

Do not plug a 5V PSU into the board yet. A known-good 5V supply into a board with shorted input caps or a fused TPS546 output will either trip the 5V brick (best case) or push current into damaged silicon and cascade the failure (worst case). Wait until cold resistance checks in Step 6 confirm the board's rails are healthy before applying any power.

Photograph the board under good light before touching it further. Phone camera, macro mode if available. Document visible damage: discolouration, craters, missing components, soot, solder-mask burns. This helps D-Central diagnose remotely if you ship the board, and it is your warranty or insurance documentation if the wrong PSU was mislabelled by its vendor.

Label every PSU in your workshop right now, before you forget which is which. `5V / 6A — Supra / Ultra / Gamma only` on one side, `12V / 5A — Hex only` on the other. Masking tape and a marker. This prevents the repeat incident that happens in roughly 30% of follow-up support tickets we see.

Visually inspect the board under magnification — a loupe or USB microscope. Focus on the TPS546 package, every ceramic cap near it, the barrel jack solder joints, the ESP32-S3 module's 3V3 LDO area, and any TVS diode or polyfuse at the input. Any crater, soot, bulge, or brown spot moves you straight to Tier 4 triage — a component has already failed.

Run cold resistance checks with the board unpowered. Multimeter on diode/continuity mode. Probe GND to the 5V input pad (healthy `>10 kOhm` initially, rising as caps charge); GND to a VCORE output cap (healthy `>1 kOhm`); GND to the ESP32-S3 3V3 pin (healthy `>5 kOhm`). Record all three. These three numbers determine which tier of salvage is even possible — a short on any one of them flags the corresponding rail as damaged.

Current-limited bench power-up is the only safe re-power path. Bench PSU at `5.0 V`, current limit `100 mA`. Connect via barrel probe or temporary pigtail. Watch current as you enable output: healthy boot shows a brief spike (over limit, PSU folds back) then settles to `50-80 mA` idle. Step the current limit up through `2 A` and then `6 A`. If current pins at the limit immediately, something is shorted — power off, do not raise the limit, return to cold resistance probing.

Replace the input TVS diode and polyfuse if present. If cold resistance shows an input-side short localised to the TVS / fuse area, this is the straightforward fix. `SMBJ6.0A` from Digi-Key costs pennies; the polyfuse is an equally simple swap. Hot-air rework (no soldering iron on surface-mount TVS). Re-verify cold resistance after the replacement before applying any power.

If input is clean but the ESP32 `3V3` rail is shorted, the onboard LDO on the ESP32-S3 module is the failure. Two paths: desolder and replace the full ESP32-S3 module (castellated pads, annoying but doable with hot air) or bypass the onboard LDO by feeding `3.3 V` externally from a separate regulator. Option one restores OEM behaviour, option two is a hack that gets you a functioning board faster if you are experienced.

Run a 24-hour burn-in after any successful repair. Board at stock frequency (`485 MHz` Gamma, `490 MHz` Supra, `485 MHz` Ultra), log hashrate and temperature every 5 minutes. Stable values across 24 hours means the board probably survived. Any drift, intermittent zero-hashrate, or thermal anomaly means silicon damage the multimeter cannot see — latent damage always eventually shows up under sustained load.

TPS546 replacement for a shorted VCORE rail. Source the TPS546D24A (Supra / Ultra older revs) or TPS546D24S (Gamma / GT / newer revs) from Digi-Key or Mouser — AxeOS `v2.7.x+` recognises both variants, older firmware recognises only one. Fine-pitch QFN hot-air rework: preheat `150 C` bottom-side, top-side hot air `310-330 C` for ~30s, no-clean flux, tweezer alignment, no soldering iron. Re-test cold resistance then attempt a current-limited bench boot.

BM1366 / BM1368 / BM1370 ASIC replacement — the hard call. Even with a fresh TPS546, if the ASIC saw overshoot voltage during VRM failure it is likely dead. BM1366 and BM1368 are scarce but available through grey-market channels; BM1370 is very hard to source and even D-Central has limited stock. BGA rework territory: pre-heat, hot air, no-clean flux, re-balling, controlled reflow. Budget 3-4 hours and `$100+ CAD` in parts — at that point consider whether a replacement board is the better call.

Replace damaged ceramic caps near the input. If Step 6 showed an input-side short that cleared after TVS replacement but visible ceramics look cracked or discoloured, swap them. Common values `10 uF / 10 V X7R 0805` and `22 uF / 10 V X7R 0805` — verify against the Bitaxe open-source schematic on GitHub for your specific rev. Failure to replace stressed-but-not-obviously-dead caps leads to intermittent failures 30-90 days later.

Inspect PCB vias and traces near the barrel jack. Severe 12V exposure can pop vias in addition to components. Strong light and loupe. A burnt via or blown trace can be repaired with a thin-gauge magnet wire bridge — ugly but functional. Damage to a main power plane, however, is usually the end of the board's economic life; a replacement board makes more sense than a multi-hour PCB salvage.

Clean the board after rework. Flux residue plus damaged components plus newly installed parts is a recipe for intermittent failures. Isopropyl `99%`, soft brush, compressed air. Do not use ultrasonic cleaning on a populated Bitaxe — the ESP32-S3 module is not ultrasonic-safe and the cleaning process can damage the module's RF shielding and crystal.

Stop DIY and ship to D-Central when: the BM1366/1368/1370 itself is visibly damaged (burns, cracks, lifted corners), multiple rails measure as shorts, you have replaced the TPS546 and the board still will not boot, or the board boots but fails burn-in. Book a Bitaxe repair slot at https://d-central.tech/services/asic-repair/. We deliver honest salvage-vs-replace calls and do not invoice jobs that are not economical.

D-Central bench process for 12V-on-5V incidents: full damage survey with scope and microscope, quantified damage report inside 48 hours, salvage quote versus replacement quote with transparent pricing, TPS546 and / or BM1366/1368/1370 rework as needed, 24-hour burn-in under sustained load, QC sheet documenting the repair. Turnaround typically `5-10` business days for salvageable boards. Canada-wide and international shipping welcomed.

Ship safely. Anti-static bag, bubble wrap, rigid outer box with at least 5 cm of foam on every side. Include the wrong or burnt PSU if you still have it — this helps us confirm the incident versus other damage vectors. One-page note: approximate 12V exposure duration, symptoms observed (smoke, popping, smell), firmware version before the incident, whether the board has been powered on since, and contact info. This saves diagnostic time at the bench and therefore saves you money on the invoice.